Beyond Lineage: A Field-Level Trust Contract for AI Data Consumers
Towards AI
•
Data Science
AI Research
They don’t tell your AI generator whether to use it. There is a moment that happens in every serious AI data pipeline project, usually a few weeks after the worked and a few weeks before the first production incident. The team has done the right things. They have column-level lineage. They know which table every field comes from, what transformation produced it, and when the pipeline last ran. Their data catalog is current, certified models are tagged, and ownership is recorded. The governance story is clean. Then an AI generator cites a gain/loss figure that covers 38% of a portfolio.